<html><body><p>a) Default the Client ID to the MAC addr<br>b) Provide an interface (rest?) on the BMC to configure / provision the Client ID to an arbitrary string (e.g. ibm-obmc -007)<br><br>The need to provision the client ID can be based on the IP addr allocation policy defined on the DHCP *server* in the data center.<br><br>One such use case view here: <a href="https://www.net.princeton.edu/announcements/dhcp-cliid-must-match-chaddr.html">https://www.net.princeton.edu/announcements/dhcp-cliid-must-match-chaddr.html</a><br><br>regards,<br>Hari !<br><br>Hariharasubramanian R.<br>Power Firmware Development<br>IBM India Systems & Technology Lab, Bangalore, India<br>Phone: +91 80 4025 6950 <br><br><img width="16" height="16" src="cid:1__=EABB0A4DDFD1AFF98f9e8a93df938690918cEAB@" border="0" alt="Inactive hide details for Patrick Williams ---03/09/2017 10:04:26 AM---All, I'm looking for feedback from any networking expert"><font color="#424282">Patrick Williams ---03/09/2017 10:04:26 AM---All, I'm looking for feedback from any networking experts. We had an issue</font><br><br><font size="2" color="#5F5F5F">From: </font><font size="2">Patrick Williams <patrick@stwcx.xyz></font><br><font size="2" color="#5F5F5F">To: </font><font size="2">OpenBMC Maillist <openbmc@lists.ozlabs.org></font><br><font size="2" color="#5F5F5F">Date: </font><font size="2">03/09/2017 10:04 AM</font><br><font size="2" color="#5F5F5F">Subject: </font><font size="2">DHCP client behavior</font><br><font size="2" color="#5F5F5F">Sent by: </font><font size="2">"openbmc" <openbmc-bounces+hramasub=in.ibm.com@lists.ozlabs.org></font><br><hr width="100%" size="2" align="left" noshade style="color:#8091A5; "><br><br><br><tt>All,<br><br>I'm looking for feedback from any networking experts. We had an issue<br>reported by Witherspoon system manufacturing that our DHCP client was<br>not sending MAC addresses as part of the DHCP request as legacy systems<br>did. I tracked this down to newer behavior in DHCP clients called<br>'DUID's, which are intended to be system-specific rather than<br>interface-specific identifiers (MACs would be interface-specific).<br><br>There are three options for IPv4 DHCP clients and two for IPv6:<br> * MAC address based client IDs (legacy behavior, IPv4 only).<br> * System UUID based DUIDs (single ID per system).<br> * IPv6 'link-layer' based DUIDs (single ID per interface?)<br><br>I'm curious if anyone has a perspective on how data-centers would like<br>a BMC to act based on the three options for IPv4. IPv6 we will need to<br>decide between system or interface based IDs. (ie. what should be our<br>default configuration; individual vendors can always change.)<br><br>I have created an issue on Github to track the long-term decision and<br>as a temporary solution I forced IPv4 to MAC.<br><br>Original issue is at [1], follow up for this assessment is at [2], "fix"<br>in Gerrit is at [3].<br><br><br>[1] </tt><tt><a href="https://github.com/openbmc/openbmc/issues/1272">https://github.com/openbmc/openbmc/issues/1272</a></tt><tt><br>[2] </tt><tt><a href="https://github.com/openbmc/openbmc/issues/1280">https://github.com/openbmc/openbmc/issues/1280</a></tt><tt><br>[3] </tt><tt><a href="https://gerrit.openbmc-project.xyz/#/c/2933/">https://gerrit.openbmc-project.xyz/#/c/2933/</a></tt><tt><br><br>-- <br>Patrick Williams<br>[attachment "signature.asc" deleted by Hariharasubramanian Ramasubramanian/India/IBM] </tt><br><br><BR>
</body></html>